Meridional Connectivity of a 25-Year Observational AMOC Record at 47°N

1. Introduction The Atlantic Meridional Overturning Circulation (AMOC) comprises a current system in the Atlantic Ocean, transporting warm and saline subtropical water northward in an upper limb and returning cold and fresh subpolar water southward in a lower limb. These currents transport large amounts of heat, making the AMOC integral to the global climate system and its past and future changes. These changes include regional temperature and precipitation (e.g., Jackson etal.,2015), tropical storm activity (e.g., Yan etal.,2017), regional sea level (e.g., Little etal.,2019) and the rate of uptake of anthropogenic carbon (e.g., Rhein etal.,2017). However, many features of the AMOC, for example, the meridional communication of variability and the location of deep water formation, are still debated (Buckley & Marshall,2016; Lozier etal.,2019). The recent IPCC report (IPCC,2021) downgraded the confidence in a weakening of the AMOC since 1850 from medium (IPCC,2019) to low due to disagreements between models and observations. These disagreements highlight the importance of long-term observations to evaluate model results. The RAPID-MOCHA-WBTS array at 26°N in the subtropical Atlantic (hereafter RAPID array), the longest observational AMOC record so far (e.g., Cunningham etal.,2007; McCarthy etal.,2015b; Moat etal.,2020), has been operational since 2004. In the center of the subpolar North Atlantic, the OSNAP array was established in 2014 (e.g., Li etal.,2021; Lozier etal.,2017). First results challenged the notion that deep convection in the Labrador Sea is the main contributor to inter-annual variability and indicated a more dominant role for the eastern Atlantic (Lozier etal.,2019). Previous studies argued that AMOC variability in the subtropics is dominated by high-frequency wind forcing, while in the subpolar North Atlantic, low-frequency buoyancy forcing is more prominent (Buckley & Marshall,2016; Jackson etal.,2022). Due to these different timescales of the forcing mechanisms, AMOC variability is expected to be coherent across latitudes on longer than decadal timescales. However, meridional Abstract Since climate model studies project a decline of the Atlantic Meridional Overturning Circulation (AMOC) in the 21st century, monitoring AMOC changes remains essential. While AMOC variability is expected to be coherent across latitudes on longer than decadal timescales, connectivity on inter-annual and seasonal timescales is less clear. Model studies and observational estimates disagree on the regions and timescales of meridional connectivity and AMOC observations at multiple latitudes are needed to study its connectivity. We calculate basin-wide AMOC volume transports (1993–2018) from measurements of the North Atlantic Changes (NOAC) array at 47°N, combining data from moored instruments with hydrography and satellite altimetry. The mean NOAC AMOC is 17.2Sv exhibiting no long-term trend. Both the unfiltered and low-pass filtered NOAC AMOC show a significant correlation with the RAPID-MOCHA-WBTS AMOC at 26°N when the NOAC AMOC leads by about one year. Plain Language Summary In the North Atlantic Ocean, currents transport heat and salt from the warm subtropical regions to the colder and less saline subpolar regions. This current system is part of the Atlantic Meridional Overturning Circulation (AMOC). This enormous northward heat transport associated with the AMOC impacts regional and global climate. In a warming world, climate models project a reduction of the AMOC, affecting, for example, air temperature and rain patterns over the continents. Continuous observations are required to investigate whether climate models realistically simulate the AMOC and assess the validity of these projections. We calculate the AMOC from moored observations, hydrography, and satellite data at 47°N from 1993 to 2018. In this period, the AMOC at 47°N does not show a trend. The 26°N AMOC lags the 47°N AMOC by about one year, indicating that the AMOC evolution at these two latitudes is connected. WETT ETAL. © 2023 The Authors. This is an open access article under the terms of the Creative Commons Attribution-NonCommercial License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited and is not used for commercial purposes. Meridional Connectivity of a 25-Year Observational AMOC Record at 47°N Simon Wett1,2 , Monika Rhein1,2 , Dagmar Kieke1,3 , Christian Mertens1 , and Martin Moritz3 1Institute of Environmental Physics, University of Bremen, Bremen, Germany, 2MARUM - Center for Marine Environmental Sciences, University of Bremen, Bremen, Germany, 3Federal Maritime and Hydrographic Agency (BSH), Hamburg, Germany Key Points: • We present a 25-year observational Atlantic Meridional Overturning Circulation (AMOC) volume transport record at 47°N with a mean basinwide AMOC transport of 17.2Sv • The inter-annual variability at 47°N is similar to the RAPID AMOC at 26°N, while the monthly variability is much stronger at 47°N • The AMOC time series at 47°N and 26°N show a significant lag correlation when the AMOC at 47°N leads by about one year Supporting Information: Supporting Information may be found in the online version of this article. Correspondence to: S. Wett, simon.we[email protected] Citation: Wett, S., Rhein, M., Kieke, D., Mertens, C., & Moritz, M. (2023). Meridional connectivity of a 25-year observational AMOC record at 47°N. Geophysical Research Letters, 50, e2023GL103284. https://doi.org/10.1029/2023GL103284 Received 13 FEB 2023 Accepted 2 AUG 2023 Author Contributions: Conceptualization: Simon Wett, Monika Rhein, Dagmar Kieke, Christian Mertens, Martin Moritz Data curation: Simon Wett, Christian Mertens, Martin Moritz Formal analysis: Simon Wett Funding acquisition: Monika Rhein, Dagmar Kieke Investigation: Simon Wett, Monika Rhein Methodology: Simon Wett, Monika Rhein, Dagmar Kieke, Christian Mertens, Martin Moritz Resources: Monika Rhein 10.1029/2023GL103284 RESEARCH LETTER 1 of 11 Geophysical Research Letters WETT ETAL. 10.1029/2023GL103284 2 of 11 connectivity on inter-annual and seasonal timescales is less clear (Bingham etal.,2007; Gu etal., 2020). Observational estimates and coupled ocean-atmosphere models disagree on the latitude ranges and timescales of meridional connectivity of the AMOC (Elipot etal.,2014; Frajka-Williams etal.,2018; Gu etal.,2020; Kostov etal.,2022). A lack of connectivity of the AMOC at two latitudes implies convergence and divergence of ocean heat and freshwater content. These redistributions impact oceanic and atmospheric dynamics, highlighting the importance of AMOC observations at different latitudes to understand the AMOC's role in the climate system. Models often indicate a breakdown of the meridional connectivity in the transition zone between the subpolar and subtropical North Atlantic (e.g., Buckley & Marshall,2016). The North Atlantic Changes (NOAC) array was established north of the subtropical gyre and south of the region where water mass transformation from the upper to the lower limb of the AMOC occurs (Rhein etal.,2011) to study the exchange between these two regimes (Figure1a). Its instrumentation was a combined effort of the University of Bremen and the Federal Maritime and Hydrographic Agency (BSH). The array was located at 47°N in the western Atlantic and followed the historical WOCE hydrographic line A02 (Kieke etal.,2009; Koltermann etal.,1999) at about 48°N in the eastern Atlantic. Transport measurements in the western basin used in this study started in 2013. In 2016, the array was extended to the eastern basin. See Moritz etal.(2021a); Nowitzki etal.(2020); Rhein etal.(2019b) for exact deployment dates. Further efforts included validating and establishing the methods and discussing the observed top-bottom transport time series of the AMOC components (Mertens etal.,2014; Moritz etal.,2021a; Nowitzki etal.,2021; Rhein etal.,2019b) and a comparison with high-resolution ocean models (Breckenfelder etal.,2017; Mertens etal.,2014). Here, we combine the individual components from these efforts to calculate the basin-wide observational AMOC volume transport of the NOAC array at 47°N. We extend them in time using satellite altimetry, yielding a 25-year AMOC time series from 1993 to 2018. To close measurement gaps, we use the observation-based ARMOR3D product. We discuss mean, trend, and variability of the NOAC AMOC and compare it to results from the RAPID array in the subtropical North Atlantic (Moat etal.,2022). 2. Data and Methods 2.1. The North Atlantic Changes (NOAC) Moored Array The NOAC array at about 47°N consisted of various moored sensors (Figure1b). In the ocean interior, we calculate geostrophic transports in eight segments from moored pressure-sensor-equipped inverted echo sounders (PIES, Nowitzki etal.,2021; Rhein etal.,2019b). At the western (Mertens etal.,2014; Rhein etal.,2019b) and eastern (Moritz etal.,2021a) boundaries, we calculate the transports from deep sea moorings equipped with acoustic current meters and profilers. Nowitzki etal.(2021), Moritz etal.(2021a), and Rhein etal.(2019b) showed that all individual flow components in the interior and the boundary currents at the NOAC array are significantly correlated with the sea surface height (SSH). Using these correlations, we extend the individual AMOC components calculated from moored instruments back to 1993 by regressing absolute transports on the satellite altimetry-derived surface velocity field (McCarthy etal.,2020; Moritz etal.,2021a; Nowitzki etal.,2021; Rhein etal.,2019b; Roessler etal.,2015). Parts of the array, especially the Canadian and European continental shelves, are not covered by moored instruments. We fill these gaps using geostrophic velocities from the ARMOR3D data set (Guinehut etal.,2012; Mulet etal.,2012). ARMOR3D comprises satellite altimetry and in situ measurements (mainly Conductivity Temperature Depth (CTD) and Argo float profiles) in weekly resolution, merged using statistical methods. It provides temperature, salinity, and geostrophic velocities on a global grid of 0.25° lateral resolution on 24 vertical levels down to 1,500m. Below 1,500m ARMOR3D consists of World Ocean Atlas climatological values. For more information, see Guinehut(2021), and Greiner etal.(2021). We use ARMOR3D velocities also for a 180kmstrip between the Eastern Boundary Current (EBC) mooring array and the nearest PIES (Figure1). Here, due to a loss of mooring instruments, it was impossible to calculate geostrophic transports from the NOAC array observations. The Ekman transports are calculated from the wind stress using the IFREMER CERSAT Global Blended Mean Wind Fields (Desbiolles etal.,2017), comprising 6-hourly global surface wind data from satellite-borne scatterometry and radiometry on a 0.25° global grid. For more information, see Bentamy et al. (2022) and Bentamy(2022). We also use this data set to link AMOC strength changes to wind stress changes. Software: Simon Wett Validation: Simon Wett, Monika Rhein Visualization: Simon Wett Writing – original draft: Simon Wett Writing – review & editing: Simon Wett, Monika Rhein, Dagmar Kieke, Christian Mertens, Martin Moritz 19448007, 2023, 16, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2023GL103284 by Cochrane Germany, Wiley Online Library on [17/02/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Geophysical Research Letters WETT ETAL. 10.1029/2023GL103284 3 of 11 2.2. AMOC Calculation We calculate the AMOC in density space as the transport between the sea surface and the density of the maximum overturning stream function. This density is obtained from the mean of 16 lowered Acoustic Doppler Current Profiler (lADCP) and CTD sections as σθ=27.67kgm −3. It is similar to the density found by Li etal.(2021) for the full OSNAP array in 2014–2018 (σθ=27.65kgm −3). At 47°N, the σθ=27.67kgm −3 isopycnal lies at about 1,100m depth in the interior, rising to 300m close to the Canadian continental slope (Figure1b). We force a net-zero top-bottom transport across the array by introducing a spatially uniform but temporally variable compensation velocity, calculated at monthly resolution and spread evenly across the whole basin, including Figure 1. (a) Schematic depiction of the major currents in the North Atlantic. Red (blue) arrows mark upper (deep) circulation pathways. Acronyms show the locations of the North Atlantic Current (NAC) and Eastern Boundary Current (EBC). Black lines indicate the transport lines of the NOAC and RAPID array. PIES locations are marked in white, moorings in orange. (b) Mean geostrophic velocity along the NOAC array, based on CTD data and ARMOR3D (on the shelves), both 2003–2020. Markers indicate the positions of each PIES (white circles), single-point acoustic current meter (CM, white squares), and upward-looking Acoustic Doppler Current Profilers (ADCP, orange). The black line shows the mean depth of the σθ=27.67kgm −3 isopycnal. Colors on top of the section indicate the geographical boundaries of the individual AMOC components (see Figure S2 in Supporting InformationS1). Topography is from ETOPO2v2 (NOAA National Geophysical Data Center, 2006). 19448007, 2023, 16, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2023GL103284 by Cochrane Germany, Wiley Online Library on [17/02/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Geophysical Research Letters WETT ETAL. 10.1029/2023GL103284 4 of 11 the shelves. For comparison with the OSNAP and RAPID estimates, other mass exchanges, for example, the Bering Strait inflow into the Arctic Ocean (0.8±0.1Sv, Woodgate & Aargaard,2005) are neglected (Kanzow etal.,2009,2010). We find a mean northward compensation transport of 3.3Sv. The basin-wide AMOC volume transport is calculated as the sum of the individual transport components of the NOAC array, the Ekman transport, and the compensation transport integrated from the surface to the density of σθ=27.67kgm −3, respectively. Unlike subpolar AMOC estimates, subtropical AMOC variability is independent of whether it is calculated in depth space or density space (Moat etal.,2020; Wang etal.,2019). Throughout this study, we compare the AMOC at 47°N, calculated in density space, to the RAPID AMOC at 26°N (Moat etal.,2022), calculated in depth space. 2.3. Statistics We present the AMOC volume transport at monthly resolution (Figure 2a) and low-pass filtered with the commonly used Butterworth filter (fourth order with a two-year cutoff period, hereafter simply low-pass filter, Figure2 a and c). Rhein etal.(2019b) and Nowitzki etal.(2021) showed that the decorrelation timescales of Figure 2. Monthly AMOC volume transport time series from 47°N (a) and 26°N (b) for the absolute transport (left y-axis) and the transport anomaly (with respect to each full-period mean, right y-axis). The dark lines represent the low-pass filtered AMOC. (c) Comparison of the low-pass filtered time series from 47°N (blue) and 26°N (brown). The dashed brown line is the altimetry-based AMOC estimate at 26°N from Sanchez-Franks etal.(2021) with the same low-pass filter applied. Note that Sanchez-Franks etal.(2021) use a different low-pass filter in their Figure 11. 19448007, 2023, 16, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2023GL103284 by Cochrane Germany, Wiley Online Library on [17/02/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Geophysical Research Letters WETT ETAL. 10.1029/2023GL103284 5 of 11 the individual AMOC components are around two months for the western and eastern transport components. We therefore calculate Pearson correlation coefficients from unfiltered and low-pass filtered time series of two-month averages for the individual AMOC components. All correlations are tested for statistical significance at the 95%-level (Section S1 in Supporting InformationS1) following Ebisuzaki(1997) (with 1,000 surrogates) and previously applied by McCarthy etal.(2015a) and Diabaté etal.(2021). Additionally, we perform a multitaper coherence analysis. We use the jLab toolbox (Lilly & Elipot,2021) to calculate the cross-spectra and calculate the squared coherence from these. 3. Results and Discussion 3.1. AMOC Mean and Variability at 47°N The mean strength of the NOAC AMOC (1993–2018: 17.2, standard error of the mean, SEM: 0.4Sv, 2004–2018: 17.5, SEM: 0.6Sv) is close to that of the RAPID AMOC (2004–2018: 16.9, SEM: 0.4Sv) and of other estimates from the subpolar region (OVIDE section (1997–2010: 18.1Sv, Mercier etal.,2015), OSNAP array (2014–2018: 16.6Sv, Li etal.,2021)). The mean NOAC AMOC in the OSNAP period (18.4Sv, Table S1 in Supporting InformationS1) is slightly larger compared to the full period but within the estimated uncertainty. The mean AMOC transports of the RAPID, NOAC, and OSNAP arrays are not significantly different within their uncertainties. Other calculations based on altimetry and Argo data (1993–2009, 41°N: 15.5Sv Willis,2010) and gridded data sets (1993–2015, 45°N: 14.3Sv Desbruyères etal.,2019) yield 10%–15% smaller mean AMOC volume transports, possibly due to methodological differences. Previous studies found no significant long-term trend in subtropical AMOC estimates after 1981 (Worthington etal.,2021, based on an empirical model). The NOAC AMOC shows a non-significant linear trend of 0.03Sv per year over the study period (standard error: 0.04Sv, representing a minimum uncertainty without accounting for auto-correlation). However, Rhein et al. (2019b) concluded that decadal trends in the individual AMOC components smaller than 10% cannot be resolved due to a possible change in the regression between SSH and transports. The overall variability of the monthly NOAC AMOC (standard deviation: 5.2Sv) is 50% higher than that of the RAPID AMOC (standard deviation: 3.5Sv) and lower than the OSNAP AMOC variability (7.6Sv, Baker etal.,2022). The different regional settings probably cause the different standard deviations at 47°N and 26°N. Stronger North Atlantic Current (NAC) variability at 47°N compared to the Florida Current at 26°N (Piecuch,2020) explains a larger NOAC AMOC variability. The standard deviations of the low-pass filtered time series (47°N: 1.7Sv, 26°N: 1.5Sv, both 2004–2018) are similar. AMOC trends might be influenced by assuming a constant density of the maximum stream function, which might vary on inter-annual timescales. Most shipboard observations were obtained in spring and summer, which might introduce a seasonal bias. The number of basin-wide sections is too low to allow for a time-dependent estimate. However, the effect would be small: a considerable density difference of, for example, Δσθ=0.05kgm −3 would result in an AMOC difference of less than 2%, which is small, compared to other transport uncertainties. The mean NOAC compensation transport (3.3Sv) lies within the range of other observational AMOC estimates from −2.9±1.8Sv at the OSNAP array (Fu & Li,2023) to approximately 10Sv at the RAPID array (Danabasoglu etal.,2021). The NOAC AMOC estimate utilizes the ARMOR3D data set to bridge gaps where geostrophic transports cannot be calculated from NOAC instruments, inducing further uncertainties. ARMOR3D only provides climatological values below 1,500 m. Thus any inter-annual variability in the deep ocean is neglected. However, we use ARMOR3D data only for the shelves (which are shallower than 1,500m) and a 180kmstrip in the eastern basin. Thus, we expect any uncertainty induced by using climatological ARMOR3D data to be small compared to other uncertainties. 3.2. The Seasonal Cycle of the AMOC at 47°N The AMOC volume transport at 47°N exhibits a significant seasonal cycle with a range of 6.0Sv at monthly resolution (Figure S1a in Supporting InformationS1), which is 35% of the AMOC's mean volume transport. However, the contribution of the seasonal cycle to the AMOC's total variability is small. Removing the seasonal cycle slightly reduces the standard deviation, from 5.2 to 4.8Sv, a decrease of 8%. The strongest AMOC volume transports appear from August to November, the weakest in January. This agrees with other observations of the 19448007, 2023, 16, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2023GL103284 by Cochrane Germany, Wiley Online Library on [17/02/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Geophysical Research Letters WETT ETAL. 10.1029/2023GL103284 6 of 11 seasonal cycle in the subpolar North Atlantic at the OSNAP array (Fu etal.,2023), finding a minimum AMOC volume transport in winter (NDJ). The shape of the seasonal cycle does not depend on whether it is calculated over the whole period, the RAPID measuring period, or the OSNAP measuring period. At shorter periods, larger uncertainties render the seasonal cycle insignificant. The seasonal cycle at the RAPID array exhibits a different shape, with a maximum in December and a minimum in May. At the RAPID array, the contribution of the seasonal cycle to the total AMOC variability is larger than at the NOAC array. For RAPID, the standard deviation of the full time series (3.5Sv) is 20% larger than without the seasonal cycle (2.8Sv). Of the individual AMOC components at 47°N (Figure S1b in Supporting InformationS1), only the NAC, shelves transport, and the Ekman transport show a significant seasonal cycle. Almost all individual AMOC components display a minimum in winter. Only the NAC shows an inverse behavior, with a maximum in December and a minimum in March, possibly related to the wind stress forcing, which exhibits a maximum in winter (not shown) and potentially influences the strength and position of the NAC via geostrophic adjustment. The wind-driven Ekman transport is in the opposite phase because it is strongest southward when the westerly winds are strongest. 3.3. Inter-Annual Variability of the NOAC and RAPID AMOC Consistent with previous model studies (e.g., Biastoch etal.,2008; Yeager & Danabasoglu,2014) and observational estimates (e.g., Jackson etal.,2019; Mercier etal.,2015), we find a decreasing NOAC AMOC in the 1990s from 19Sv in 1994 to 15Sv in 2000 (Figures2a and2c, low-pass filtered). This decline ceases after 2002, when the NOAC AMOC increases, reaching the maximum of more than 20Sv in late 2005. This maximum results from a strong NAC, not compensated by an average interior transport (Figure S2 in Supporting InformationS1, Section3.4). Following the 2005 maximum, the NOAC AMOC decreases with a minimum of less than 15Sv reached in 2009. It strengthens afterward until 2018. Comparable multiannual features are also observed in the RAPID AMOC: A maximum around 2005/2006 followed by a decline until about 2010, and a strengthening until 2018 (Moat etal.,2020). Though the unfiltered and low-pass filtered NOAC and RAPID AMOC time series are not significantly correlated at zero lag, the correlation becomes statistically significant (r=0.70−0.83, depending on the filtering method) when the NOAC AMOC leads by about one year (Figure S3 and Table S3 in Supporting InformationS1). This significant correlation also holds for the unfiltered time series of two-month averages. We find a significant correlation larger than r=0.30 when the NOAC AMOC leads by 12–14months (Figure S3 in Supporting InformationS1), persisting when the seasonal signal is removed from both time series (r=0.35, 12months lag). While the values of the correlation coefficients differ slightly, the statistical significance does not depend on the choice of the low-pass filter (Table S3 in Supporting InformationS1). Sanchez-Franks etal.(2021) extended the RAPID AMOC time series using satellite altimetry. Consistent with the NOAC AMOC, their time series exhibits an increase in 2000–2005, though their AMOC estimate shows no decline in the 1990s (Figure2c). We compare the NOAC AMOC time series to this estimate and find no significant correlation at any lag (Figure S3 in Supporting InformationS1). At the 90% significance level, however, the correlation at about one year lag is significant. The absence of a significant correlation possibly originates from the low correlation before the RAPID measuring period. While different altimetry-based estimates of the 26°N AMOC correlate well with the RAPID measurements (Frajka-Williams,2015; Sanchez-Franks etal.,2021), they show little agreement among themselves before, possibly due to overfitting during the RAPID array measuring period (Sanchez-Franks etal.,2021). For the RAPID measuring period (2004–2018), we find a significant lag correlation between the NOAC AMOC and the estimate by Sanchez-Franks etal.(2021) when the 47°N AMOC leads by about one year (r=0.68). 3.4. Main AMOC Components At the RAPID array, the mid-ocean transport dominates the low-frequency (>1year) variability of the AMOC (Moat etal., 2020). For the NOAC AMOC, the NAC and the interior transport dominate the variability on all timescales (Figures S2 and S5a in Supporting InformationS1). Of all components, the NAC exhibits the strongest correlation with the basin-wide NOAC AMOC (r=0.61, Table S2 in Supporting InformationS1). 19448007, 2023, 16, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2023GL103284 by Cochrane Germany, Wiley Online Library on [17/02/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Geophysical Research Letters WETT ETAL. 10.1029/2023GL103284 7 of 11 In the subtropical North Atlantic, on sub-annual timescales, the transport through Florida Strait and the upper mid-ocean transport compensate each other (Frajka-Williams etal.,2016; Kanzow etal.,2007). Consistent with Rhein etal.(2019b) and Mertens etal.(2014), we find a similar relationship at 47°N between the corresponding components of the NAC and the interior transport. The net southward interior transport exhibits a mean of −6.6Sv. The interior transport is significantly anticorrelated with the NOAC AMOC (r=−0.41, Table S2 in Supporting InformationS1) and the NAC (r=−0.81). The shelf transport is concentrated primarily over the western shelf with a mean southward transport of −11.2Sv. The Western Boundary Current (WBC), Eastern Boundary Current (EBC), and the Ekman transport have mean values below 5Sv and standard deviations of 2–4Sv, with strong sub-annual variability. The NAC dominates the upper transports in large parts of the western basin, including the region of the WBC component (Figure1), resulting in a mean northward flow of the WBC component above σθ=27.67kgm −3. 3.5. Atmospheric Drivers of AMOC Variability Processes influencing the AMOC on multiple timescales include advection, direct wind forcing, geostrophic adjustments to wind forcing by waves, or surface buoyancy fluxes (e.g., Buckley & Marshall,2016; Kostov etal.,2021; Kostov etal.,2022; Robson etal.,2016; Wang etal.,2015). Despite the limited influence of the wind-driven Ekman transport on the AMOC (Figure S2 in Supporting InformationS1), inter-annual AMOC variability in the subpolar North Atlantic partly results from wind stress anomalies (e.g., Buckley & Marshall,2016; Jackson et al., 2022; Kostov et al., 2021). At 47°N, strong AMOC phases, for example, around 1993/1994, 2005/2006, and 2015/2016 (visible in the low-pass filtered time series, Figures 2a and2c) are accompanied by a strong NAC (Figure S2 in Supporting InformationS1), a relationship also found by other studies (Smeed etal.,2014,2018). These phases coincide with sharp transitions in the local wind stress anomaly (Figure S4 in Supporting InformationS1), possibly linked to changes in atmospheric pressure patterns resulting in NAO anomalies or meridional shifts of the pressure systems (Buckley & Marshall,2016; Iqbal etal.,2019). Recalculating the NOAC AMOC with one component held constant at its mean value allows us to investigate their contribution to the total AMOC. In these strong AMOC phases, the AMOC with a constant NAC is about half as strong as with a varying NAC (not shown), indicating that a strong NAC has driven these strong AMOC phases. The contribution of the Ekman transport to the AMOC is not exceptionally strong in these phases. We conclude that large-scale wind-field changes in the North Atlantic have partly driven the strong NOAC AMOC periods around 1993/1994, 2005/2006, and 2015/2016, possibly via changes in the geostrophic adjustment, resulting in a strong NAC that additionally may have shifted zonally. The primary non-seasonal atmospheric mode in the North Atlantic is the North Atlantic Oscillation (NAO), affecting wind stress and surface buoyancy fluxes. We use the monthly mean NAO index from the U.S. National Oceanic and Atmospheric Administration based on orthogonally rotated principal components (Barnston & Livezey,1987). The correlation between the low-pass filtered NOAC AMOC and the NAO index with the same filter applied is insignificant. However, the positive NAO phase from 2014 to 2018 coincides with an AMOC increase of about 1Sv at the RAPID array. At the NOAC array, the increase in this period is even more substantial (over 2Sv), suggesting that a long-lasting positive NAO phase drives an increased AMOC volume transport. This also holds for the strong NOAC AMOC phase in 1993/1994, a period of persistent positive NAO (Buckley & Marshall,2016). 3.6. Meridional Connectivity Between the AMOC at 47°N and 26°N The significant correlation between the low-pass filtered NOAC AMOC and the RAPID AMOC with a lag of about one year possibly indicates a southward communication of transport signals (e.g., Biastoch etal.,2008; Zhang,2010). Although advective southward communication is possible on these timescales, the lag correlations of the southward AMOC transport components at 47°N (interior transports, shelves, Ekman transport) with the RAPID AMOC (not shown) are statistically insignificant. Thus, we cannot attribute the southward communication of anomalies to a single component. Possibly the southward communication is maintained by other mechanisms, for example, large-scale atmospheric dynamics or advection in the lower AMOC limb (Zou etal.,2019). Wind-stress variations drive coherent AMOC variability across latitudes on inter-annual timescales (Elipot etal.,2017). The wind forcing results in the Ekman transport, oriented northward at 26°N and southward at 47°N. 19448007, 2023, 16, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2023GL103284 by Cochrane Germany, Wiley Online Library on [17/02/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Geophysical Research Letters WETT ETAL. 10.1029/2023GL103284 8 of 11 This inverse behavior could reduce the correlation between the NOAC AMOC and the RAPID AMOC. However, the Ekman transport additionally influences the AMOC via the compensation transport. To eliminate the direct influence of the Ekman transport on meridional connectivity, we compute the correlation between the NOAC AMOC and the RAPID AMOC with the NOAC Ekman component held constant at its mean value. Consistent with model studies (Bingham etal.,2007; Wang etal.,2019), we find an increased and significant correlation between the low-pass filtered NOAC AMOC and the RAPID AMOC at zero lag (r=0.66), indicating that meridional connectivity is maintained by other processes, possibly still due to wind forcing, for example, geostrophic adjustment to large-scale wind field changes (Elipot etal.,2017). While a lag correlation analysis focuses on the time domain of the connection between the signals, a coherence analysis provides insights into the shared behavior of the time series in the frequency domain. Using a multitaper estimate (Figure S6 in Supporting InformationS1), we find coherence between the two signals, exceeding the 95% confidence level at periods of about seven years. Here, the phase lag is about 50°, indicating a time lag of about one year. This signal might correspond to the decrease following the 2005/2006 maximum and the subsequent increase after 2010, visible in both AMOC time series. However, the time series length limits the analysis, and the coherence might extend to lower frequencies, not resolved due to the limited overlap of the time series. The number of frequency bands exceeding the 95% confidence level is too low to exclude statistically that any coherence occurs by chance. However, our correlation analysis shows that the time series are related. 4. Conclusions We present a 25-year (1993–2018) observational AMOC record based on the NOAC array at 47°N in the North Atlantic. The mean and inter-annual standard deviation of the NOAC AMOC agree well with estimates at the subtropical RAPID array at 26°N, while variability on sub-annual timescales is larger at the NOAC array compared to the RAPID array. As discussed in Jackson etal.(2022), the inter-annual variability of AMOC transport time series at 41°N (Willis,2010), 45°N (Desbruyères etal.,2019), 50°N (Jackson etal.,2019), and the OVIDE section (Mercier etal.,2015) differ substantially. These differences could originate from the different methods and data sets used. All these estimates suffer from undersampling of the strong and narrow boundary currents, especially at the western continental slope and rise (e.g., Mertens etal.,2014), while the NOAC array at 47°N was specifically designed to resolve the boundary currents. Jackson etal.(2022) noted the correlation between the AMOC time series at 26°N and 41°N (Willis,2010), and associated both with the subtropical regime. The authors assigned the AMOC estimates from 45°N (Desbruyères etal.,2019) and 50°N (Jackson etal.,2019) to the subpolar regime because they did not appear to be correlated with the subtropical AMOC. The NOAC and RAPID AMOC, however, exhibit a significant correlation when the NOAC AMOC leads by about one year. This finding either points to an extended subtropical regime reaching as far north as 47°N or challenges the notion that the subtropical and subpolar regimes are decoupled on inter-annual and longer timescales. We find increased coherence of the two time series at periods of about seven years, but the number of frequency bands exceeding the 95% confidence level is too low to exclude random coherence. The present time series are too short to resolve coherence on decadal or longer timescales. It remains an open question what determines the meridional connectivity between 47°N and 26°N. The lag correlations of the individual components of the NOAC AMOC with the RAPID AMOC remain insignificant. The higher correlation between both time series when the 47°N Ekman transport variability is removed from the NOAC AMOC indicates the potential impact of other wind-driven mechanisms, for example, geostrophic adjustment. Furthermore, the question of how far the meridional connectivity extends north beyond 47°N remains to be answered. Addressing this question requires more efforts to separate regional from methodological differences in the AMOC time series (Danabasoglu et al., 2021; Frajka-Williams et al., 2018) and a deeper understanding of the involved processes. Both objectives highlight the importance of long-term observations, especially AMOC time series and near-realistic high-resolution coupled models. The OSNAP array in the center of the subpolar gyre will provide a valuable tool to study the meridional connectivity into the central subpolar gyre once its AMOC time series is sufficiently long. Data Availability Statement The basin-wide AMOC volume transport time series from the NOAC array at 47°N, displayed in Figure 2, is freely available from PANGAEA under the CC-BY license (Wett et al., 2023, https://doi.org/10.1594/ PANGAEA.959558). Top-bottom PIES transport time series from the eastern basin, published in Nowitzki etal.(2021) are available from PANGAEA (Nowitzki etal.,2020, https://doi.org/10.1594/PANGAEA.925089). 19448007, 2023, 16, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2023GL103284 by Cochrane Germany, Wiley Online Library on [17/02/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License Geophysical Research Letters WETT ETAL. 10.1029/2023GL103284 9 of 11 Mooring current meter data and transport time series, published in Moritz etal. (2021a) are available from PANGAEA (Moritz etal.,2021b, https://doi.org/10.1594/PANGAEA.932566). Lowered ADCP, current meter data, and PIES time series measurements from the western basin, published in Rhein etal.(2019b) are available from PANGAEA (Rhein etal.,2019a, https://doi.org/10.1594/PANGAEA.903211). This study has been conducted using E.U. Copernicus Marine Service Information; ARMOR3D: https://www.doi.org/10.48670/ moi-00052, product ID: MULTIOBS_GLO_PHY_TSUV_3D_MYNRT_015_012 (downloaded on May 20/21, 2021) and IFREMER CERSAT Global Blended Mean Wind Fields: Product ID: WIND_GLO_WIND_L4_REP_ OBSERVATIONS_012_006 (downloaded in August 2022). Data from the RAPID AMOC monitoring project (Moat et al., 2022, https://www.doi.org/10.5285/e91b10af-6f0a-7fa7-e053-6c86abc05a09) is funded by the Natural Environment Research Council and freely available from www.rapid.ac.uk/rapidmoc (downloaded on 25 January 2023). The monthly mean NAO index data is available at https://www.cpc.ncep.noaa.gov/products/ precip/CWlink/pna/nao.shtml#current (downloaded on 15 December 2022). The topography in Figure 1 is from ETOPO2v2 (NOAA National Geophysical Data Center, 2006, downloaded on 4 November 2020). References Baker, J., Renshaw, R., Jackson, L., Dubois, C., Iovino, D., & Zuo, H. (2022). Overturning variations in the subpolar North Atlantic in an ocean reanalyses ensemble. Journal of Operational Oceanography, Copernicus Marine Service Ocean State Report, (6), 16–20. https://doi.org/10. 1080/1755876X.2022.2095169 Barnston, A. G., & Livezey, R. E. (1987). Classification, seasonality and persistence of low-frequency atmospheric circulation patterns. Monthly Weather Review, 115(6), 1083–1126. https://doi.org/10.1175/1520-0493(1987)115<1083:CSAPOL>2.0.CO;2 Bentamy, A. (2022). Quality information document for the global ocean wind products WIND_GLO_WIND_L4_REP_OBSERVATIONS_012_006. Copernicus Marine Service. Bentamy, A., Piollé, J. F., Prevost, C., & Giesen, R. (2022). Product user manual for wind product WIND_GLO_WIND_L4_REP_OBSERVATIONS_012_006. Copernicus Marine Service. Biastoch, A., Böning, C. W., Getzlaff, J., Molines, J.-M., & Madec, G. (2008). Causes of interannual–decadal variability in the Meridional Overturning Circulation of the midlatitude North Atlantic Ocean. Journal of Climate, 21(24), 6599–6615. https://doi.org/10.1175/2008JCLI2404.1 Bingham, R. J., Hughes, C. W., Roussenov, V., & Williams, R. G. (2007). Meridional coherence of the North Atlantic Meridional Overturning Circulation. Geophysical Research Letters, 34(23), L23606. https://doi.org/10.1029/2007GL031731 Breckenfelder, T., Rhein, M., Roessler, A., Böning, C. W., Biastoch, A., Behrens, E., & Mertens, C. (2017). Flow paths and variability of the North Atlantic current: A comparison of observations and a high-resolution model. Journal of Geophysical Research: Oceans, 122(4), 2686–2708. https://doi.org/10.1002/2016JC012444 Buckley, M. W., & Marshall, J. (2016). Observations, inferences, and mechanisms of the Atlantic Meridional Overturning Circulation: A review. Reviews of Geophysics, 54(1), 5–63. https://doi.org/10.1002/2015RG000493 Cunningham, S. A., Kanzow, T., Rayner, D., Baringer, M. O., Johns, W. E., Marotzke, J., etal. (2007). Temporal variability of the Atlantic Meridional Overturning Circulation at 26.5°N. Science, 317(5840), 935–938. https://doi.org/10.1126/science.1141304 Danabasoglu, G., Castruccio, F. S., Small, R. J., Tomas, R., Frajka-Williams, E., & Lankhorst, M. (2021). Revisiting AMOC transport estimates from observations and models. Geophysical Research Letters, 48(10), e2021GL093045. https://doi.org/10.1029/2021GL093045 Desbiolles, F., Bentamy, A., Blanke, B., Roy, C., Mestas-Nuñez, A. M., Grodsky, S. A., etal. (2017). Two decades [1992–2012] of surface wind analyses based on satellite scatterometer observations. Journal of Marine Systems, 168, 38–56. https://doi.org/10.1016/j.jmarsys.2017.01.003 Desbruyères, D. G., Mercier, H., Maze, G., & Daniault, N. (2019). Surface predictor of overturning circulation and heat content change in the subpolar North Atlantic. Ocean Science, 15(3), 809–817. https://doi.org/10.5194/os-15-809-2019 Diabaté, S. T., Swingedouw, D., Hirschi, J. J.-M., Duchez, A., Leadbitter, P. J., Haigh, I. D., & McCarthy, G. D. (2021). Western boundary circulation and coastal sea-level variability in northern hemisphere oceans. Ocean Science, 17(5), 1449–1471. https://doi.org/10.5194/ os-17-1449-2021 Ebisuzaki, W. (1997). A method to estimate the statistical significance of a correlation when the data are serially correlated. Journal of Climate, 10(9), 2147–2153. https://doi.org/10.1175/1520-0442(1997)010<2147:AMTETS>2.0.CO;2 Elipot, S., Frajka-Williams, E., Hughes, C. W., Olhede, S., & Lankhorst, M. (2017). Observed basin-scale response of the North Atlantic Meridional Overturning Circulation to wind stress forcing. Journal of Climate, 30(6), 2029–2054. https://doi.org/10.1175/JCLI-D-16-0664.1 Elipot, S., Frajka-Williams, E., Hughes, C. W., & Willis, J. K. (2014). The observed North Atlantic Meridional Overturning Circulation: Its meridional coherence and ocean bottom pressure. Journal of Physical Oceanography, 44(2), 517–537. https://doi.org/10.1175/JPO-D-13-026.1 Frajka-Williams, E. (2015). Estimating the Atlantic overturning at 26°N using satellite altimetry and cable measurements. Geophysical Research Letters, 42(9), 3458–3464. https://doi.org/10.1002/2015GL063220 Frajka-Williams, E., Lankhorst, M., Koelling, J., & Send, U. (2018). Coherent circulation changes in the deep North Atlantic from 16°N and 26°N transport arrays. Journal of Geophysical Research: Oceans, 123(5), 3427–3443. https://doi.org/10.1029/2018JC013949 Frajka-Williams, E., Meinen, C. S., Johns, W. E., Smeed, D. A., Duchez, A., Lawrence, A. J., etal. (2016). Compensation between meridional flow components of the Atlantic MOC at 26° N. Ocean Science, 12(2), 481–493. https://doi.org/10.5194/os-12-481-2016 Fu, Y., & Li, F. (2023). OSNAP technical report, Calculation method summary. Retrieved from https://duke.app.box.com/v/2023OSNAP technicalreport Fu, Y., Lozier, M. S., Biló, T. C., Bower, A. S., Cunningham, S. A., Cyr, F., etal. (2023). Seasonality of the Meridional Overturning Circulation in the subpolar North Atlantic. Communications Earth & Environment, 4(1), 181. https://doi.org/10.1038/s43247-023-00848-9 Greiner, E., Verbrugge, N., Mulet, S., & Guinehut, S. (2021). Multi observation global ocean 3D temperature salinity heights geostrophic currents and MLD product MULTIOBS_GLO_PHY_TSUV_3D_MYNRT_015_012. Copernicus Marine Service. Gu, S., Liu, Z., & Wu, L. (2020). Time scale dependence of the meridional coherence of the Atlantic Meridional Overturning Circulation. Journal of Geophysical Research: Oceans, 125(3), e2019JC015838. https://doi.org/10.1029/2019JC015838 Guinehut, S. (2021). Multi observation global ocean 3D temperature salinity heights geostrophic currents and MLD product MULTIOBS_GLO_ PHY_TSUV_3D_MYNRT_015_012. Copernicus Marine Service. Acknowledgments S. Wett was funded by the German Science Foundation (DFG) through the International Research Training Group ArcTrain “Processes and impacts of climate change in the North Atlantic Ocean and the Canadian Arctic” (IRTG 1904 ArcTrain to M. Rhein). Shipboard and moored data at 47°N—the NOAC array—have been funded by the German Ministry of Education and Research (Grants 03F0443C, 03F0605C, 03F0561C to M. Rhein, and 03F0792A to M. Rhein and D. Kieke). M. Rhein and C. Mertens received funding by the EU Project 101059547—EPOC. EPOC is funded by the European Union. Views and opinions expressed are however those of the authors only and do not necessarily reflect those of the European Union. Neither the European Union nor the granting authority can be held responsible for them. The authors thank H. Nowitzki (formerly University of Bremen) for preparing data from the eastern PIES, R. Steinfeldt (University of Bremen) for calibrating the NOAC CTD data, A. Sanchez-Franks (National Oceanography Centre, Southampton) for supplying the extended transport time series at 26°N and L. Caesar (University of Bremen) for helpful comments on the manuscript. Additionally, the authors thank the editor K. Karnauskas and the reviewers S. Elipot and an anonymous reviewer for a thorough and very constructive review of the manuscript. Open Access funding enabled and organized by Projekt DEAL. 19448007, 2023, 16, Downloaded from https://agupubs.onlinelibrary.wiley.com/doi/10.1029/2023GL103284 by Cochrane Germany, Wiley Online Library on [17/02/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License